Author 
Message 
Arby science forum beginner
Joined: 19 Jul 2006
Posts: 1

Posted: Wed Jul 19, 2006 12:09 pm Post subject:
modelling method for categorical data



Hello all,
This is my first post to this list, my question is actually about
statistics although I'm no statistician so if I don't get the correct
terminology then forgive me but I'll try. What I want to know is, what
would be the right type of statistical modelling for
nominal/categorical data? Put another way what is the appropriate
method for predicting a nominal response variable from a number of
categorical explanatory variables?
I am actually a biologist and what I have is a dataset that records
gains and losses of regions of DNA from 400 colorectal cancer patients.
More specifically, the whole human genome is divided into 862 segments,
each segment is scored as 0 = no change, 1 = loss, 1 =gain, 2 = high
level amplification, for each of the 400 patients.
In general terms my question would be: can I predict the status of 1 of
the 862 segments if I know the status of the other 861 segments? That
would be what I think is called the maximal model, what is the process
for reducing this to the minimal number of segments necessary to
predict a given segment with a set level of certainty? Having looked
through some basic stats books the nearest thing I've found is ANOVA
but this confused as it talked about calculating means and I don't see
how you can have a mean of a nominal variable (this may be the crux of
my problem).
I'm not looking for complete solutions, I just don't have the correct
vocabulary to describe what I need so I'd be grateful if someone could
tell me what the appropriate terms are and hopefully provide some
beginners references so I can go away and learn how to do it.
regards,
Richard 

Back to top 


Stephen J. Herschkorn science forum Guru
Joined: 24 Mar 2005
Posts: 641

Posted: Wed Jul 19, 2006 3:56 pm Post subject:
Re: modelling method for categorical data



In sci.math, Arby wrote:
Quote:  Hello all,
This is my first post to this list, my question is actually about
statistics although I'm no statistician so if I don't get the correct
terminology then forgive me but I'll try. What I want to know is, what
would be the right type of statistical modelling for
nominal/categorical data? Put another way what is the appropriate
method for predicting a nominal response variable from a number of
categorical explanatory variables?
I am actually a biologist and what I have is a dataset that records
gains and losses of regions of DNA from 400 colorectal cancer patients.
More specifically, the whole human genome is divided into 862 segments,
each segment is scored as 0 = no change, 1 = loss, 1 =gain, 2 = high
level amplification, for each of the 400 patients.
In general terms my question would be: can I predict the status of 1 of
the 862 segments if I know the status of the other 861 segments? That
would be what I think is called the maximal model, what is the process
for reducing this to the minimal number of segments necessary to
predict a given segment with a set level of certainty? Having looked
through some basic stats books the nearest thing I've found is ANOVA
but this confused as it talked about calculating means and I don't see
how you can have a mean of a nominal variable (this may be the crux of
my problem).
I'm not looking for complete solutions, I just don't have the correct
vocabulary to describe what I need so I'd be grateful if someone could
tell me what the appropriate terms are and hopefully provide some
beginners references so I can go away and learn how to do it.
regards,
Richard

This post is more germane to the newsgroup sci.stat, where I have
crossposted this reply.

Stephen J. Herschkorn sjherschko@netscape.net\
Math Tutor on the Internet and in Central New Jersey and Manhattan 

Back to top 


Google


Back to top 



The time now is Tue Oct 23, 2018 1:18 am  All times are GMT

